PINN-kit¶
A comprehensive toolkit for Physics-Informed Neural Networks (PINNs) that empowers researchers and engineers to solve complex differential equations using deep learning.
!!! abstract "What is PINN-kit?" PINN-kit provides an intuitive interface for implementing and training physics-informed neural networks, combining the power of deep learning with the rigor of physics-based modeling.
Key Features¶
- Easy-to-use interface for defining physics-informed neural networks
- Flexible domain handling utilities for arbitrary input variables
- High-performance training with PyTorch backend
- Built-in visualization tools for results analysis
- Support for various differential equation types
- Clean, modular design for easy customization
Quick Start¶
Installation¶
PINN-kit supports macOS, Linux, and Windows. Install it using pip:
!!! tip "Virtual Environment" It's recommended to use a virtual environment with Python 3.12+ to avoid dependency conflicts:
```bash
python3.12 -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
pip install pinn-kit
```
Basic Usage¶
from pinn_kit import PINN, Domain
# Create a domain with flexible variable definition
domain = Domain([
('x', -1, 1), # x-coordinate bounds
('y', -1, 1), # y-coordinate bounds
('t', 0, 1) # time bounds
])
# Initialize a PINN with 3 inputs, 2 hidden layers, 1 output
pinn = PINN([3, 20, 20, 1])
# Define your physics-informed loss function
def physics_loss():
# Your physics constraints here
return loss_value
# Train the network
pinn.train_model(domain, physics_loss, epochs=1000)
!!! success "That's it!" PINN-kit handles all the complexity behind the scenes, letting you focus on your physics problem.
Documentation¶
Explore our comprehensive documentation to get the most out of PINN-kit:
Learn how to use PINN-kit effectively with step-by-step tutorials and examples.
Understand how to define and work with different types of domains and boundaries.
Dive deep into the PINN implementation and advanced configuration options.
See real-world applications and use cases with complete code examples.
Use Cases¶
PINN-kit is perfect for solving:
- Heat transfer and fluid dynamics problems
- Wave propagation and vibration analysis
- Quantum mechanics and electromagnetic field simulations
- Financial modeling with differential constraints
- Biological systems and chemical reactions
System Requirements¶
| Component | Requirement |
|---|---|
| Python | 3.9+ |
| PyTorch | 2.0+ |
| NumPy | 1.21+ |
| OS | macOS, Linux, Windows |
!!! note "Dependencies" All required dependencies (including PyTorch) are automatically installed with PINN-kit.
Contributing¶
We welcome contributions! Whether it's:
- Bug reports and feature requests
- Documentation improvements
- Code contributions and pull requests
- Ideas and suggestions
Visit our GitHub repository to get started.
License¶
This project is licensed under the MIT License - see the LICENSE file for details.